Precision = the fraction of retrieved items that are relevant
(How much of what you retrieved is good?)
# relevant articles
# total articles
Recall = fraction of relevant items retrieved out of all relevant items available in the database
(How much of the good stuff did you actually get. Unfortunately, the higher the recall, the more 'junk' you end up getting also.)
# relevant articles retrieved
# of total relevant articles available
When searching, you're looking for a reasonable balance between precision (narrowing your search to get ONLY relevant articles) and recall (widening it to get ALL relevant articles, which usually means a lot more junk to weed through as well).
A common question is "How many articles should be retrieved by a good search?" There's no exact answer to that. Somewhere between 100-300 is a reasonable number of abstracts to weed through, but it depends greatly on your question, how comprehensive you want to be, and how much literature there truly is on your topic.
22 y/o female
Gestation = 23rd week
Presented to the ED with 3 days of right lower quadrant pain, anorexia, and persistent nausea/vomiting
WBC = 12,000/mm3
Hemoglobin = 12.1
Hematocrit = 34.9%
Platelet = 306,000/mm3
Meds = prenatal vitamins, folate, and iron sulfate
DX = acute appendicitis
Weight: 98 kg
Mallampati Score: 2
Full range of neck motion
Thyromental space > 5 centimeters
Smaller than normal endotracheal tube used
Question: Which induction agent should be used?
Now let's practice parsing the question into terms of PICO(T)
How effective is the prophylactic use of ondansetron for preventing postoperative nausea and vomiting in adult surgical patients?
1. Parse the question into concepts
2. List terms for each concept including both textwords and MeSH subject headings
3. Use terms to create search strategies and run them in PubMed
4. Look at results
a. How many abstracts did you get?
b. How can you increase the yield?
c. How can you decrease the yield?
d. How relevant are your results